facebook_wav2vec2-large_October_03_2023_05h34PM

This model is a fine-tuned version of facebook/wav2vec2-large on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.1318
• Accuracy: 0.9743

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.4214	1.0	121	0.3650	0.7932
0.1959	2.0	242	0.2588	0.8960
0.1365	3.0	363	0.0732	0.9713
0.1003	4.0	484	0.0849	0.9719
0.0806	5.0	605	0.2170	0.9381
0.0588	6.0	726	0.0738	0.9760
0.0472	7.0	847	0.2083	0.9409
0.0505	8.0	968	0.1020	0.9760
0.0427	9.0	1089	0.0626	0.9809
0.0285	10.0	1210	0.1062	0.9732
0.0286	11.0	1331	0.1068	0.9738
0.0231	12.0	1452	0.1137	0.9755
0.0232	13.0	1573	0.0783	0.9815
0.0158	14.0	1694	0.1138	0.9755
0.0164	15.0	1815	0.1318	0.9743

Framework versions

• Transformers 4.33.3
• Pytorch 2.2.0.dev20230927+cu121
• Datasets 2.14.5
• Tokenizers 0.13.3